Multi-dimensional adherence patterns in rare haematological conditions: Longitudinal analysis of real-world patient-reported outcomes

Abstract Background Capturing longitudinal patient-reported outcomes (PROs) beyond healthcare visits is central to understanding real-world disease burden and therapeutic response. However, availability of high-adherence long-term follow-up (FU) data remains a challenge. Sustainable engagement is critical to generating translatable evidence of patient experience over time. Aims To assess variation in digital PRO adherence rates across subgroups (age, sex, disease, geography, and FU) and explore the feasibility of sustained remote engagement to inform future strategies that support care models and intervention monitoring. Methods Data from patients enrolled in a digital ecosystem with PRO-capture via disease-specific apps were analysed across multiple conditions (255 UK and 51 Indian sickle cell disease (SCD), 43 Waldenström's macroglobulinaemia (WM), 261 myeloproliferative neoplasms (MPN), 15 multiple myeloma (MM), and 5 chronic myeloid leukaemia (CML)) using the EQ-5D-5L as a cross-disease metric. Patients were onboarded on a rolling basis (June 2021-July 2025). WM and SCD-UK received PRO incentives (daily voucher points). MPN was part of an ongoing research study (IRAS 332286). The remaining cohorts were real-world and non-incentivised. All groups received fortnightly check-ins from a Good Clinical Practice (GCP)-certified team to support technology use. No in-app prompts or push notifications were used. Two cohorts were defined. Cohort 1 (n=631) included all patients with ≥1 PRO and ≥30 days FU to reflect overall adherence. Cohort 2 (n=261) comprised a subcohort with ≥1 PRO within 30 days prior to data cut-off and ≥90 days FU, to reflect sustained engagement. Weekly adherence was defined as ≥1 PRO/week. Analyses were stratified by disease, age, sex, and FU. Results The mean (SE) age, proportion female, and mean (SE) FU in weeks for Cohort 1 were 35 (1), 73%, 149 (2) in SCD-UK; 27 (1), 45%, 51 (2) in SCD-India; 59 (1), 64%, 44 (1) in MPN; 66 (2), 54%, 78 (2) in WM; 54 (3), 64%, 19 (1) in MM; and 64 (5), 60%, 36 (1) in CML. Full cohort linear correlation analysis found increasing age associated with higher adherence, while longer FU correlated with lower adherence (both p<0.001). The highest mean weekly adherence rates were seen in WM (77%), MM (76%) and MPN (72%), with lower rates in CML (55%), SCD-UK (33%), and SCD-India (29%). Adherence patterns did not follow a clear trend by study or incentivisation status, with intermediate adherence in the study cohort (MPN) and low adherence in the incentivised SCD-UK group, in contrast to high adherence in incentivised WM. Adherence by age and FU group matched full cohort correlation patterns across all diseases. Of note, even the least adherent age and FU groups across all conditions (SCD-India, under 25: 26%; 6-12 months: 23%) equated to approximately one datapoint per four weeks. While no sex-based disparity was seen in SCD-UK, adherence was higher in females than males in MM (37%), CML (15%), SCD-India (9%), MPN (7%), and WM (4%). Similar to Cohort 1, analysis of the longitudinal high-engagement subcohort (Cohort 2) saw significant correlations between increasing adherence and increasing age (p<0.001). Mean disease-level adherence rates were 74% in SCD-UK (n=52), 72% in SCD-India (n=7), 90% in MPN (n=164), 96% in WM (n=28), 97% in MM (n=8), and 65% in CML (n=2). Adherence by FU group remained stable in Cohort 2. In SCD-UK, with the longest mean FU across conditions, adherence was 69% (1-2 years), 80% (2-3 years), and 70% (3+ years). Conclusion Long-term digital PRO adherence was feasible across all ages, with higher rates in older cohorts, challenging assumptions around digital disutility in ageing populations. While lower adherence in SCD may reflect younger age, disease severity may also contribute, with periods of intense pain and hospitalisation impacting patients' ability to engage regularly. Similar rates in UK and Indian SCD groups support cross-regional applicability. Mean rates in even the least adherent groups equated to approximately one datapoint per four weeks, exceeding typical healthcare database frequency. In Cohort 2, declines over time were mitigated, highlighting the value of embedding long-term habits to sustain high-adherence capture. Importantly, incentives alone were insufficient to drive adherence in high-burden groups. Future work will explore drivers of these adherence patterns to inform targeted support strategies, particularly for younger populations.